Recent searches
No recent searches
![Erin O'Callaghan's Avatar](https://support.zendesk.com/system/photos/1900042430764/Profile_Photo_Updated___Cropped.png)
Erin O'Callaghan
Joined Oct 16, 2021
·
Last activity Jan 24, 2025
Following
0
Followers
11
Total activity
493
Votes
11
Subscriptions
181
ACTIVITY OVERVIEW
BADGES
ARTICLES
POSTS
COMMUNITY COMMENTS
ARTICLE COMMENTS
ACTIVITY OVERVIEW
Latest activity by Erin O'Callaghan
Erin O'Callaghan created an article,
If you’ve created an AI agent for messaging, there are several standard responses you can customize to define the AI agent's default responses to customer input during a live conversation:
- Start of the conversation: Initial greeting when a messaging conversation is started.
- After the AI agent replies: Response to customer feedback.
- Escalation: What happens when the customer wants to talk to a human.
- If the AI agent can't answer the question: Response when the customer's comment does not match any answer, known as a fallback response.
This article contains the following sections:
- Viewing the standard responses
- Customizing the greeting message
- Customizing the response to customer feedback
- Customizing the escalation response
- Customizing the fallback response
Related articles:
Viewing the standard responses
The AI agent has several standard responses that you can customize as needed.
To view the AI agent standard responses
- In Admin Center, click
Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage AI agents for messaging.
- Click the name of the AI agent you want to edit.
- Click the Behavior tab, then click each message to view it:
Customizing the greeting message
When your customers first open a messaging conversation in the Web Widget or mobile SDK, they’re met with an initial greeting message. You can update the greeting message or leave it as-is.
To customize the greeting message
- On the Behavior tab, expand the Start of the conversation section.
- In the Greeting field, enter your initial message or leave the default message. (Hi there. Got a question? I'm here to help.)
- Select Add option to talk to a human if you want a button labeled
“Talk to a human” to appear after the greeting message.
If a customer clicks “Talk to a human,” the escalation response is triggered.
The response is saved automatically as you edit it, but updates will not be presented to customers until you publish the AI agent.
Customizing the response to customer feedback
When the AI agent presents an answer, it follows up by asking whether the answer was helpful. The "After the AI agent replies" response lets you configure the AI agent response when an end user says the answer was helpful or unhelpful.
To customize the response to customer feedback
- On the Behavior tab, expand the After the AI agent replies section.
- On the Helpful answer tab, in the Message field, enter a message to display when the end user says the suggested article was helpful, or use the default message provided. (Great. You can ask another question anytime.)
- Click Add option to talk to a human if you want a button labeled
“Talk to a human” to appear after the AI agent’s response to customer
feedback.
If a customer clicks “Talk to a human,” the escalation response is triggered.
- Click the Unhelpful answer tab and configure the options as described above.
The response is saved automatically as you edit it, but updates will not be presented to customers until you publish the AI agent.
Customizing the escalation response
The escalation response determines what happens when the customer wants to talk to a human. This response is triggered either when:
- The AI agent determines that a human agent is needed in order to solve the customer’s request.
- The customer clicks the “Talk to a human” button that appears after the AI agent’s response. This button appears if you configure the greeting message, response to customer feedback, or fallback response to include it.
To customize the escalation response
- On the Behavior tab, expand the Escalation section.
- Under the Business hours heading, use the drop-down to select a
schedule to set the AI agent’s business hours, or select Always
online.
For more information about schedules, see Setting your schedule with business hours and holidays.
- Under the Responses heading, on the During business hours tab,
select one of the following options:
-
Create a ticket: Select this option if you want a ticket to
be created when the customer escalates an AI agent conversation to a
human.
If you select this option, configure the following fields:
- Message: Enter a message asking the customer to provide details about their request.
- Customer details: Select the fields a customer should fill out when escalating.
- Follow-up message: Enter a message telling the customer what to expect after they’ve provided their details.
-
Show a custom response: Select this option if you don’t want
a ticket to create a ticket when a customer escalates an AI agent
conversation to a human, and instead show the customer a pre-written
message. If you select this option, configure the following fields:
- Message: Enter a message directing the customers to a website, form, email address, or telephone number.
-
Create a ticket: Select this option if you want a ticket to
be created when the customer escalates an AI agent conversation to a
human.
- Click the Outside business hours tab and configure the options as described above.
The response is saved automatically as you edit it, but updates will not be presented to customers until you publish the AI agent.
Customizing the fallback response
The fallback response is triggered when there is no answer that matches the end user's question or comment.
To customize the fallback response
- On the Behavior tab, expand the If the AI agent can't understand the question section.
- Enter a message the AI agent will send if it can’t provide an answer based
on the information the customer entered.
The default fallback response is, Sorry I couldn’t answer your question.
- Click Add option to talk to a human if you want a button labeled
“Talk to a human” to appear after the fallback response.
If a customer clicks “Talk to a human,” the escalation response is triggered.
The response is saved automatically as you edit it, but updates will not be presented to customers until you publish the AI agent.
Edited Feb 03, 2025 · Erin O'Callaghan
0
Followers
1
Vote
0
Comments
Erin O'Callaghan created an article,
AI agent personas determine the style of expression applied to AI-generated messages, giving them a consistent voice that reflects your brand. You configure an AI agent persona as part of an AI agent’s general settings.
This article includes the following sections:
Related articles:
About AI agent personas
AI agent personas let you apply a personality to an AI agent’s responses. A persona is composed of three settings that work together to determine how your AI agent presents itself when interacting with customers:
- Business profile: A short description of what your company does. This information gives the AI agent context so that it can give the most appropriate responses.
- Tone of voice: The general style of language the AI agent should use when responding to customers.
- Reply length: The approximate length of each of the AI agent’s responses.
Configuring an AI agent persona
The option to configure an AI agent persona appears in your AI agent’s general settings.
To configure a persona
- In Admin Center, click
Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage AI agents for messaging.
- Click the AI agent you want to configure a persona for.
- Click the Settings tab.
- Expand the Persona section.
- Under Business profile, enter one or two short, factual sentences in English that describe what your company does. Keep this description focused on simple text about your company's business domain, rather than marketing-focused material that could potentially influence your AI agent’s behavior.
- Under Tone of voice, select one of the following options:
- Professional: (Default) Polite and direct
- Informal: Casual and friendly
- Enthusiastic: Upbeat and friendly
- Under Reply length, select one of the following options:
- Short: 30–60 words
- Medium: 60–90 words
- Long: 90–120 words
- Very long: 120–150 words
Your changes are saved automatically as you make them, but updates to the AI agent’s persona won’t be presented to customers until you publish the AI agent.
Edited Feb 03, 2025 · Erin O'Callaghan
0
Followers
1
Vote
0
Comments
Erin O'Callaghan created an article,
AI agents - Advanced is an add-on available for purchase for Suite and Support plans that enhances the AI agent functionality in Zendesk.
This article answers the following questions:
- What is AI agents - Advanced?
- How does AI agents - Advanced work?
- Why should I use AI agents - Advanced?
- Who can use AI agents - Advanced?
- How do I turn on AI agents - Advanced?
- Where can I learn more?
Related articles:
What is AI agents - Advanced?
The AI agents - Advanced add-on is an upgrade to Zendesk’s AI agent functionality. The add-on includes all the features of the AI agents - Essential level, in addition to advanced capabilities such as conversation flows that combine generative and scripted responses, comprehensive API access, and advanced analytics.
How does AI agents - Advanced work?
With AI agents - Advanced, you can automate more complex customer queries, create scripted conversation flows for specific scenarios, integrate with other systems to fully automate requests, and benefit from in-depth reporting on your AI agent usage and automation rates.
The following are key features of AI agents - Advanced:
- Zero-training AI agents: These advanced AI agents use generative AI to understand a customer's message and either answer directly based on your help center content, or else link to an appropriate conversation flow designed by your admins.
- Dialogue builder: This tool lets you create complex, automated conversation flows, including hybrid flows that combine generative AI responses with scripted flows.
- Integration builder: This tool lets you leverage comprehensive API access and orchestration to integrate your AI agents with your customer relationship management (CRM) software and other third-party tools.
- Advanced analytics: A comprehensive AI agent performance dashboard gives you insights into key automation metrics, lets you report on the value of AI agent outcomes, and helps you make data-driven decisions to improve future efficiency.
For even more about what AI agents - Advanced contains, see Navigating AI agents - Advanced functionality.
Why should I use AI agents - Advanced?
You should consider upgrading to AI agents - Advanced if you have high support volumes or complex automation requirements. These features are specifically designed to significantly increase your automation rates, even for complex customer service scenarios.
Who can use AI agents - Advanced?
The AI agents - Advanced add-on is available for purchase for customers on any Suite or Support plan.
How do I turn on AI agents - Advanced?
First, you must purchase the AI agents - Advanced add-on. To do so, contact Zendesk Sales. After you’ve purchased the add-on, see Getting started with AI agents - Advanced.
Where can I learn more?
For more information, see AI agent resources.
Edited Feb 03, 2025 · Erin O'Callaghan
0
Followers
1
Vote
0
Comments
Erin O'Callaghan created an article,
AI agents are the next generation of AI-powered bots that automate and resolve your customers’ issues across service channels. AI agent functionality is split into multiple levels, each offering the support you need for different business requirements.
AI agents - Advanced is available as an add-on for any Suite and Support plans, and includes all features of the AI agents - Essential level in addition to advanced functionality such as conversation flows, APIs, and advanced analytics.
This article gives you an overview of how to start using the most important functionality included in AI agents - Advanced, including zero-training AI agents, the dialogue builder, the integration builder, and the performance overview dashboard.
This article contains the following topics:
- Getting started with zero-training AI agents
- Getting started with conversation flows for AI agents
- Getting started with integrating external systems with AI agents
Related articles:
Getting started with zero-training AI agents
Zero-training AI agents use generative AI to understand a customer's message and either answer directly based on your help center content, or else link to an appropriate conversation flow designed by your admins. These AI agents are quick to set up and easy to maintain.
To learn more about zero-training agents, see About zero-training AI agents.
To get started creating a zero-training AI agent, you’ll need to perform the following main tasks:
Create a zero-training AI agent
Creating a zero-training AI agent is the first step in automating customer service solutions for your customers. Zero-training AI agents use generative AI-powered use cases to understand a customer's message and link to the appropriate dialogues.
For step-by-step setup instructions, see Creating a zero-training AI agent.
Configure the AI agent's settings
AI agent settings control many things, including how and where your AI agent appears to customers, what persona and tone of voice it uses in communicating with them, the languages it’s available in, and the conversation flows it follows.
For more information about the available settings for an AI agent, see Settings explained.
Create use cases
Use cases are the mechanism by which zero-training AI agents understand what a customer is asking about and connect them with the right dialogue. Use cases leverage generative AI to power AI agents without the time-consuming activity of training the AI agent. Use cases simplify setup and maintenance while enhancing the AI agent's understanding and performance.
For step-by-step instructions, see Creating and managing use cases for zero-training AI agents.
Create dialogues
The dialogue builder is the tool that lets you create conversation flows for your AI agents. With it, you can create everything from simple to complex flows—and even interact with external systems—to resolve customer requests without a human agent having to get involved.
For step-by-step instructions, see Using the dialogue builder to create conversation flows for AI agents.
Monitor AI agent performance
The final and ongoing step in creating a zero-training AI agent is to monitor its performance and continue to improve its efficiency. To do so, you can use the Performance Overview dashboard, which gives you insights into key automation metrics, lets you report on the value of AI agent outcomes, and helps you make data-driven decisions to improve future efficiency.
For details, see Analyzing AI agents - Advanced with the Performance Overview dashboard.
Getting started with conversation flows for AI agents
The dialogue builder lets admins create complex, automated conversation flows, including hybrid flows that combine generative AI responses with scripted flows.
If you followed the instructions in the Create a zero-training AI agent section above, you’ve already used the dialogue builder to create dialogues for your zero-training AI agent.
For more information about what you can do with the dialogue builder, see Building dialogues for AI agents - Advanced.
Getting started with integrating external systems with AI agents
The integration builder lets you leverage comprehensive API access and orchestration to integrate your AI agents with your customer relationship management (CRM) software and other third-party tools.
To get started with integrating AI agents - Advanced with any of the following CRMs, see the linked resources:
- Integrating Sunshine Conversations with AI agents - Advanced
- Integrating Zendesk Support with AI agents - Advanced
- Integrating Zendesk Chat with AI agents - Advanced
- Integrating Salesforce with AI agents - Advanced
- Integrating Freshworks with AI agents - Advanced
- Integrating other platforms with AI agents - Advanced
Edited Feb 11, 2025 · Erin O'Callaghan
0
Followers
2
Votes
0
Comments
Erin O'Callaghan created an article,
Instructions and custom tone of voice let you control AI agent behavior in ways beyond the default persona options. This article discusses best practices for creating effective instructions and custom tone of voice, and gives you examples of good and bad instructions.
The following video gives you an overview of creating instructions and custom tone of voice:
Deploying instructions and custom tone of voice for AI agents (3:57)
This article contains the following topics:
- Best practices for creating effective instructions
- Examples of good and bad instructions
- Best practices for customizing a tone of voice
- Example of a custom tone of voice
Related articles:
- Using instructions to influence advanced AI agent responses
- Customizing the persona and tone of voice for your advanced AI agent
Best practices for creating effective instructions
To get the best results from your instructions, keep the following best practices in mind:
- Remember the purpose of instructions
- Be clear and direct
- Use capitalization for emphasis
- Create one instruction per directive
- Avoid conflicting instructions
- Use existing features appropriately
- Always test your instructions
Remember the purpose of instructions
Instructions allow you to refine an AI agent's responses beyond just tone of voice. They can enforce specific language, avoid certain phrases, or ensure key information is included in replies.
What instructions can’t do, however, is fundamentally change the content returned by UltimateGPT. For example, instructions cannot be used to search a different knowledge source or to cause an escalation.
Instructions are applied after UltimateGPT has understood the end user’s message, queried a knowledge source accordingly, and generated an answer ready to be sent back to the end user. At this point in the process, instructions can be used to shape—but not fundamentally change—the way that answer is presented to the end user.
Be clear and direct
When crafting instructions, clarity is key. Use straightforward language and direct commands.
For example, instead of saying:
- "I want the AI to always use our brand name."
Say this instead:
- "ALWAYS use 'Acme Labs™' instead of 'Acme.'"
A direct approach helps the AI understand and implement your instructions more effectively.
That said, expect your instructions and tone of voice settings to be interpreted as strong guidance rather than rigid commands. Due to the way generative AI works, your AI agents might not execute every instruction perfectly every time, so it’s important to maintain realistic expectations about their performance.
Use capitalization for emphasis
When you want to stress certain behaviors in your instructions, using capitalization can be effective.
For example, instructing the AI agent to "ALWAYS include a link" will highlight the importance of that directive.
Create one instruction per directive
Instead of including multiple directives in a single instruction, break them down into separate instructions. Ensure that each instruction contains no more than a single directive. This approach:
- Enhances the likelihood that the AI agent will respect your instructions.
- Makes it easier for you to edit and manage instructions over time.
Avoid conflicting instructions
Ensure that your instructions don’t contradict each other. For example, don’t write an instruction that says to end all messages with a smiley face if you have another instruction that says not to use emojis.
Also, make sure your instructions don’t conflict with your chosen tone of voice. For example, if you configured the AI agent’s tone of voice to be friendly and casual, avoid instructions that suggest overly formal or abrupt responses. Consistency in messaging is crucial for a seamless customer experience.
Use existing features appropriately
Remember that instructions are not a substitute for other features. For specific tasks like defining a welcome message, setting the default tone of voice, or setting pronoun formality for supported languages, use the designated features instead of relying on instructions.
In particular, take full advantage of the available persona and tone of voice settings. These features help define how your AI agent interacts with customers, enhancing the overall user experience.
Always test your instructions
Before making any instruction active, test the individual instruction and test all your instructions together to see how the AI agent responds. Testing helps you identify any issues with wording or clarity. If an instruction doesn’t yield the expected outcome, try rephrasing it. The wording of an instruction plays a significant role in determining how an AI agent responds.
After deploying your AI agent, continue to monitor its interactions with end users. If certain instructions aren’t performing as expected, revisit and tweak them. The AI agent’s effectiveness can improve over time with careful adjustments based on real-world usage.
Examples of good and bad instructions
Below are examples of instructions, some good and some bad. Each example also includes an analysis of how following or ignoring the best practices above impact the instruction’s effectiveness.
Good instructions
-
Example: If the user asks any questions about how we are regulated, then send the following response verbatim: “Thank you for reaching out to us. Please see our regulatory information at https://help-centre-link.com/regulation.”
- Why it’s good: Clear and direct.
-
Example: ALWAYS say “Enjoy your flight!” at the end of your message
- Why it’s good: Uses capitalization for emphasis.
-
Example: When your answer is a list of step-by-step instructions, always list them as 1, 2, 3 regardless of the formatting of the source.
- Why it’s good: Clear and direct.
-
Example: At FinanceCo, we cannot offer any financial advice ever. Make sure you NEVER give the user financial advice of any kind. If you think an answer could be construed as financial advice, make sure to add a disclaimer at the beginning that says “Please note, we cannot provide financial advice.” followed by a new paragraph.
- Why it’s good: Uses capitalization for emphasis. Though note that, as mentioned above, instructions cannot be 100% foolproof, so expect some potential instances of the AI agent not including the disclaimer when you might want it to.
Bad instructions
-
Example: You can use emojis, but only if they are more likely to be appropriate than not. Decide yourself.
- Why it’s bad: The language is too vague and leaves too much up to interpretation.
-
Example: Always welcome the user.
- Why it’s bad: This conflicts with the welcome message feature.
-
Example: Always escalate to an agent if the user mentions being hacked.
- Why it’s bad: Instructions cannot trigger escalations. This should be handled via a use case instead.
Best practices for customizing a tone of voice
You can also consider a custom tone of voice to be a kind of instruction, just directed at one specific aspect of your AI agent. You'll see best results if your custom tone of voice is as descriptive as possible, potentially with suggestions of the types of phrases or emojis that the AI agent should use.
As with instructions, be direct about what the AI agent should do. Consider a wide range of scenarios and whether the tone of voice will suit them all. For example, an AI agent that has been instructed to always be cheery and lighthearted might not provide the best answer to a user who is complaining they've been doxxed.
Example of a custom tone of voice
Below is an example of a custom tone of voice that is descriptive and contains examples of specific phrases and emojis.
Example: Respond to the user like Cher from the movie Clueless. Be super bubbly, confident, and a total valley girl. Use phrases like “As if!”, “Whatever!”, and “Totally!” Give responses in a fun, sassy, and fashionable way, with lots of enthusiasm and charm! Use one or two emojis per message.
Edited Feb 11, 2025 · Erin O'Callaghan
1
Follower
2
Votes
1
Comment
Erin O'Callaghan created an article,
Intelligent triage leverages AI to automatically classify incoming tickets by intent, language and sentiment. The Zendesk Intent Model includes pre-trained intents across several industries, providing relevant intents and use cases tailored to each account's ticket data.
To give admins greater flexibility over the available intents, you can create custom intents that address issues specific to your business needs. This improves automation accuracy and helps build confidence in automated decisions.
This article contains the following topics:
Related articles:
About custom intents
Previously, you could request a new intent, a process that required human evaluation for an intent to be added to your account. With this EAP, we're eliminating the wait time entirely. Custom intents can now be created instantly, added immediately to your list of intents, and made available right away to start detecting the intent of incoming customer requests.
As such, the request new intent feature is retired for any accounts enrolled in this EAP.
Custom intents are specific to each Zendesk account, meaning any intents you create are not available to other Zendesk accounts.
Custom intents are not intended to identify details of the ticket, such as specific product or service names, branch locations, subscription types, or similar details. Instead, see the entity detection feature to identify business-critical information in your incoming requests automatically.
Requirements
To create custom intents, you must:
- Have the Advanced AI add-on and have an intent model assigned.
- Sign up for the EAP.
- Have intents turned on.
Limitations
The following limitations apply to custom intents:
- You can create up to 50 custom intents. Deactivated intents don’t count towards this limit.
- English is the only supported language. Custom intent detections are made only for tickets with content in English.
- Custom intents aren't compared against existing Zendesk intents or other custom intents to check for duplicated, ambiguous, or conflicting intents. The quality of custom intents depends on the quality of the name, description, and ticket examples you provide when creating the intent.
- AI agents cannot detect custom intents.
Creating a custom intent
Admins can create custom intents in Admin Center. A custom intent should explain the main reason the end user is reaching out. It should cover a single reason.
Before creating a custom intent, check whether a similar intent already exists. Similar or overlapping intents can lead to low-confidence or wrong intent detection.
To create a custom intent
- In Admin Center, click
Objects and rules in the sidebar, then select Business rules > Intelligent triage.
- Click Intent.
- Select the Intent list tab.
- Click the Actions drop-down and select Create custom intent.
The Create custom intent page opens. - Fill in the following required fields:
- Name: A short, descriptive name for the intent. The intent name is visible to agents, so make sure it’s clear and concise so they can quickly understand it.
- Description: An explanation of what the ticket is about. Clearly define what the intent should include and exclude. Write the description as if you’re explaining an issue to an agent on their first day. Consider including examples and common words that users might use.
- Category: The category and subcategory that the new intent applies to. Intents are organized hierarchically in three levels: category, subcategory, and intent. The category is the broad reason for an incoming request, the subcategory provides a more detailed breakdown of the reason, and the intent is a single, very specific reason.
-
Tickets with this intent: Example tickets that the new intent
applies to. You must list up to ten examples, and each ticket must:
- Have the full URL (for example, https://.zendesk.com/agent/tickets/1)
- Come from the Email channel
- Be in English
- Have a subject and description
- Be unique (in other words, tickets can't be repeated) but must represent the same intent
- Click Create custom intent.
The new intent is added to the Intent list tab.
Examples of custom intents
Below are some examples of well-defined custom intents:
Product not working as expected
- Name: Product not working as expected
- Category: Order
- Sub-category: Product issue
- Description: The requester wants to report a product that is not functioning correctly, has developed a fault, or is not meeting quality expectations.
Cancel direct debit
- Name: Cancel direct debit
- Category: Billing
- Sub-category: Change payment method
- Description: The customer wants to stop or cancel an existing direct debit or automatic payment arrangement. They may cite reasons such as financial constraints, changes in circumstances, or dissatisfaction with the service.
Change booking arrival location
- Name: Change booking arrival location
- Category: Travel
- Sub-category: Booking update
- Description: The end user wants to modify the arrival location of their booking. This could include changing flight destinations, correcting errors in the booked location, or updating hotel addresses.
Deactivating a custom intent
You can deactivate custom intents that are no longer needed by your account. When you deactivate a custom intent, the intent will no longer be detected in new tickets. If you have any business rules, such as triggers, view or automations, that are based on the deactivated intent, they will stop working.
You can also activate inactive custom intents as needed. When you activate a custom intent, the intent is detected in new tickets and any business rules that are based on the intent work as they did previously.
Keep in mind that your account has a limit of 50 active custom intents. If you’ve reached that limit and try to activate an inactive custom intent, a message notifies you that the limit has been reached and the custom intent can’t be activated.
To activate or deactivate a custom intent
- In Admin Center, click
Objects and rules in the sidebar, then select Business rules > Intelligent triage.
- Click Intent.
- Select the Intent list tab.
- Browse or search to find the custom intent you want to activate or deactivate.
- Hover your mouse over the intent, click the options (
) icon, and select Activate or Deactivate.
- Confirm your selection by clicking either Activate intent or
Deactivate intent.
The intent’s status is updated in the intent list. Deactivated intents are labeled “Inactive”. If you’ve activated an intent, the “Inactive” label is removed.
Edited Feb 12, 2025 · Erin O'Callaghan
0
Followers
6
Votes
3
Comments
Erin O'Callaghan commented,
Hi Bruno Gabriel, unfortunately it's not currently possible to show exactly which agents are online or offline without first drilling into the live status metric.
View comment · Posted Jan 07, 2025 · Erin O'Callaghan
0
Followers
0
Votes
0
Comments
Erin O'Callaghan created an article,
It’s common for agents to use Zendesk to interact with end users, and then use Jira to create issues when help is required from the IT or engineering team.
In this recipe, we’ll streamline that process by creating an auto assist procedure to guide agents through troubleshooting common problems and automatically creating a Jira issue for the IT team if more specialized assistance is required.
This article contains the following topics:
- Workflow goal
- Prerequisites
- Creating a Jira project
- Creating a Jira app and OAuth client
- Getting the client ID and secret
- Finding your Jira Cloud ID
- Registering your Jira OAuth client in Zendesk
- Creating a connection between Zendesk and Jira
- Creating an action to create a Jira issue
- Creating a procedure to guide agents
- Testing your workflow
- Customizing your procedure
- Troubleshooting issues
Related articles:
Workflow goal
We’re going to use auto assist to help agents troubleshoot common IT issues, and, if necessary, create a Jira issue to request help from the IT team.
To facilitate this, we’ll do the following:
- In Jira:
- Identify or create a new Jira project, so we have somewhere to create Jira issues.
- Create a Jira OAuth client.
- In Zendesk:
- Register the Jira OAuth client so you can use it to create an OAuth token.
- Create a connection, which makes it easy to create an OAuth token and safely store it for use with your action.
- Create an action, which uses the connection to make a call to the Jira API and create a Jira issue.
- Create a procedure, which tells auto assist what steps are required to guide the agent through creating a Jira issue.
- Test that everything is working.
Prerequisites
To accomplish this workflow, you’ll need the following:
- A Jira Cloud instance Tip: Consider using a new Jira sandbox or free plan instance rather than your production Jira instance until you’re comfortable with how everything works.
- A Jira user with sufficient permissions to create a project, create issues
within that project, and create a Jira OAuth client.Tip: If you’re using a Jira Free plan, all users are admins and have the permissions you need.
- A Zendesk account with the Advanced AI add-on.
- An admin user login for the Zendesk account.
Creating a Jira project
In Jira, identify a project to use or create a new one. It doesn’t matter whether it’s scrum or kanban. Make sure your Jira user has access to the project.
Make note of the key for your project (for example, ENGREQ), as you’ll need this later.
Creating a Jira app and OAuth client
To create a Jira application and OAuth 2.0 integration, perform the steps described in Enabling OAuth 2.0 (3LO) in the Jira Cloud Platform documentation. Make sure to configure the following values.
https://zis.zendesk.com/api/services/zis/connections/oauth/callback
When adding an API to the app:
- Select Permissions in the left menu.
- Find Jira API and click Add.
After a moment, the Add button will change to Configure.
- Click Configure.
- Find Jira platform REST API and click Edit Scopes.
- Make sure View Jira issue data and Create and manage issues
are selected.
This will allow the action you create later to create, read, and change Jira issues.
- Click Save.
Getting the client ID and secret
In Jira, get the client ID and secret:
- Open the Jira Developer Console.
- Go to the Settings tab.
- Make a note of the Client ID and Secret.
Finding your Jira Cloud ID
https://.atlassian.net/_edge/tenant_info
Make a note of the Cloud ID, as you’ll need it when you create the action later.
Registering your Jira OAuth client in Zendesk
Registering your Jira OAuth client allows you to use it within Zendesk, instead of having to interact with it using API requests.
To register your Jira OAuth client in Zendesk
- In Admin Center, click
Apps and integrations in the sidebar, then select Connections > OAuth Clients.
- Click Add client.
- Fill out the following fields:
- Name: Give the client a meaningful name so you can find it easily later (for example, Jira_client).
- OAuth grant type: Authorization code
- Client ID: The Jira client's ID
- Client Secret: The Jira client's secret
- Authorize URL: auth.atlassian.com/authorize Token URL: auth.atlassian.com/oauth/token
- Scopes: read:jira-work write:jira-work offline_access
- Click Save.
Creating a connection between Zendesk and Jira
To create the connection
- In Admin Center, click
Apps and integrations in the sidebar, then select Connections > Connections.
- Click Create connection.
- Fill out the following fields:
- Connection name: Give the client a meaningful name so you can easily find it later (for example, Jira_oauth_token).
- Choose authentication type: OAuth 2.0
- Client: Select the Jira OAuth client you registered in Zendesk.
- Scopes: Leave this blank.
- Allowed domain: api.atlassian.com
- Click Save.
You’re taken through an OAuth screen to allow Zendesk to access your Jira instance.
Tip: If you don’t see this screen, check that it hasn’t been blocked by your browser. - Click Accept.
You’re returned to the Connections page, where you should see a new entry for your new connection. If the Create connection screen you were working on is still open, you can safely close it now.
Creating an action to create a Jira issue
In Zendesk, you’ll create an action that uses the Jira issue API endpoint to create a new issue.
To create the action
-
In Admin Center, click
Apps and integrations in the sidebar, then select Actions and webhooks > Actions.
- Click Create action.
- In the Name field, enter Create IT task.
- In the Description field, enter Creates an IT task for the team to do, and responds with the ID of the task.
- Define the following three Inputs:
Input name Input description Input type name A brief name for the task. string description Full details of the task. Must be less than 200 characters. string priority Priority indicates how severe or urgent the task is. Priority is represented by a single number. Valid values are typically 2 (high impact or urgency issue), 3 (standard issue), and 4 (low impact or urgency issue). string Note: You may need to modify the numeric values used for the priority to suit your specific Jira setup. - Define your API configuration:
- Request method: POST
- Endpoint URL: https://api.atlassian.com/ex/jira//rest/api/2/issue
-
Authentication: Select the connection you created
earlier.
- Define the Body:
{ "fields": { "summary": "{{name}}", "description": "{{description}}", "priority": { "id": "{{priority}}" }, "issuetype": { "id": "10004" }, "project": { "key": "ENQREQ" } } }
- Use the {+} icon to insert the placeholders for priority, name, and description.
- The project key value should be the key you identified earlier.
- 10004 is the default ID for the Jira issue type Task. Depending on
your Jira setup, you might need to use a different ID. Learn how to
find your issue type IDs in Jira here.
- Define an Output:
Output name Output description task_id The human-readable ID for the new task. - Click Save.
Creating a procedure to guide agents
To create a procedure
- In Guide admin, click Procedures in the sidebar.
- Click Create procedure.
- In the Name field, enter a descriptive name, such as IT task helper.
- In the body field, enter the following content:
Your job is to help create IT tasks for the IT team.
Step 1: Ask the user what they need help with.- List the following common requests:
- Order new access card
- Drink tray will not open
- Printer problem
- Other
Step 2: Gather more details about the task.- Order new access card: Ask the user to provide name, position description, eye color.
- Drink tray will not open: Ask the user if they have tried pressing the button on the front of the PC tower. Ask if they have tried right-click and eject.
- Printer problem: Ask for the printer name, which is located on the front of the unit. Ask for more details about the problem.
- Other: This is a catchall for any other issue. Ask the user to explain what the problem is.
Step 3: Determine priority.Step 4: Summarize issue and get user confirmation.- Ask the user what impact this is having on them and how urgently it needs to be fixed.
- Based on this information, assign a priority of 2 (high), 3 (regular), or 4 (low).
- Based on the information provided, generate a name and description for the task. The description must be no longer than 200 characters.
- Show the name and description to the user. Ask them to confirm.
- Strictly do not create the task until the user has confirmed.
Step 5: Create IT task.- Tell the user you are going to create the task now and to please wait a moment. For example, "Thanks for confirming the details. I'll log the task with the team now. Please hold on a moment."
- Then create an IT task using the name, description, and priority generated in steps 3 and 4.
Step 6: Share IT task ID.- After the IT task is created, give the task ID to the user.
Step 7: Wrap up- Thank the user and wrap up the conversation.
- List the following common requests:
- Click Save.
Testing your workflow
The simplest way to test your workflow is using the test feature in a conversation bot.
To test your workflow in a conversation bot
- In a conversation bot, open or create an answer that includes the Transfer to agent step type.
- On the Transfer to agent step, add the tag agent_copilot_enabled.
- On the edit page for your conversation bot, click Test bot.
For help, see Testing a conversation bot before publication.
- Start a conversation with the bot using a phrase related to the answer you
created.
In this browser window, you’ll play the role of the end user.
- In another browser window, open the Zendesk Agent Workspace.
Here, you’ll play the role of the agent. You should see a notification pop up with the new conversation.
- As the agent, follow the auto assist suggestions to reply to the end user.
- As the end user, respond with the information requested.
- When the Jira issue is created, open the link to see the new issue.
Customizing your procedure
You can further customize the procedure and action in this workflow recipe to suit your needs. For example:
- In steps 1 and 2 of the procedure, you could change the list of common tasks and troubleshooting steps to suit your specific requirements.
- In step 3 of the procedure, you could add different priority levels. If you do, make sure you also update the input in the action you created.
- You could add additional steps to gather further information that gets added to corresponding fields on the Jira Issue (for example, the environment the issue is happening in, or data for custom fields). If you do, make sure you also update the inputs and body in the action you created.
You can also create a similar “Can I get an update on my IT task?” procedure so that end users can stay informed about the progress of their ticket. To get started:
- Create a new action that fetches a Jira Issue based on the issue ID and returns useful information like the status and last updated time as outputs.
- Create a new procedure that asks the user for their issue ID and gives them the latest update on the issue.
Troubleshooting issues
If you encounter authorization problems with Jira:
- Check that your Jira user has the right permissions to access the project.
If you encounter issues with the action:
- Check the action executions in the integration log.
- Make sure the numeric values in the action align with your Jira setup (for example, priority 2/3/4, issue type 10004) and that you’re using the correct key for your project (for example, ENGREQ).
If the procedure is not working as expected:
- Check that the naming and description for the action and its inputs and outputs are consistent with the naming you use in the procedure.
If auto assist doesn’t start on the ticket:
- Check that the Zendesk user you’re testing with is included in the groups that have access to auto assist. See Turning on and configuring access to auto assist.
- Check that your conversation bot answer includes the tag agent_copilot_enabled.
Edited Feb 03, 2025 · Erin O'Callaghan
0
Followers
2
Votes
0
Comments
Erin O'Callaghan commented,
Zach Gilbert Yes, happy to clarify! There's an optional step during auto assist setup where you can add a tag to specify which procedure you've created that auto assist should follow for a given scenario:
- For adding tags to conversation bot answers, the optional step is here.
- For adding tags using triggers, the optional step is here.
I've updated this recipe to link to that information so it's easier to find!
Auto assist can't currently apply ticket tags as an action, but according to this comment from our Product Manager, that enhancement is being considered for the future.
View comment · Posted Dec 20, 2024 · Erin O'Callaghan
0
Followers
0
Votes
0
Comments
Erin O'Callaghan created an article,
After you create actions to allow AI agents to perform certain tasks during conversations with end users, you can review, edit, or delete those actions as needed.
This article contains the following topics:
Related articles:
- About actions for advanced AI agents
- Creating and adding actions for advanced AI agents
- About events for advanced AI agents
Reviewing actions in conversation logs
In conversation logs, you can find detailed information about which actions were applied to a conversation.
To review actions in conversations logs
- In the AI agents - Advanced add-on, in the top-right corner, use the AI agent drop-down field to select the AI agent you want to review actions for.
- In the main menu on the left, select Conversation logs.
- Select a conversation.
In the conversation pane that appears, any actions executed during the conversation appear in gray boxes. - Select an action in the conversation to see which action was executed and at what level (AI agent, intent, or block).Note: Actions that were applied to a conversation before the introduction of reusable actions don’t have a snapshot view. Clicking them takes you directly to where they were applied.
- Click the action name to open the Executed action details pane.
This pane shows the configuration of the action as it was when it was executed during the conversation, even if the action’s configuration has since changed.
- (Optional) Click View action to open the Action details pane.
From this pane, you can:
- Click Edit action to update the action’s configuration on the Actions page.
- Select any of the entries under the Usage heading to jump to the action’s configuration at the AI agent level, intent level, or block level.
Editing an action
Because actions can be reused at multiple levels, easy management is crucial for keeping your actions up to date. When you update an action, those updates are applied to every location where the action is used.
Actions can be edited from the Actions page, or from any place where actions can be added.
To edit an action
- In the AI agents - Advanced add-on, in the top-right corner, use the AI agent drop-down field to select the AI agent you want to edit actions for.
- In the main menu on the left, select Content > Actions.
- Select the action you want to edit.
The Action details pane appears. Under Usage, note the locations where the action is currently used. All of these locations will be affected by any updates you make to the action. - Click Edit action.
- Update the action’s configuration as necessary.
Tip: For help filling out the fields in the Create action dialog, see Creating an action. - Click Apply to all instances.
The number included in the button’s label tells you how many locations will be affected by your changes.
Your changes are applied to the action going forward. Any instances of the action that were already executed in conversations are not affected.
Deleting an action
Deleting an action affects all the locations where it’s used. Ensure that you no longer need the action anywhere before deleting it.
If an action is still needed in some areas but not in others, remove it from the specific locations where it should no longer be triggered instead of deleting it.
To delete an action
- In the AI agents - Advanced add-on, in the top-right corner, use the AI agent drop-down field to select the AI agent you want to delete actions for.
- In the main menu on the left, select Content > Actions.
Hover your mouse over the action you want to delete and select the options () menu.
- Select Delete.
The Delete action dialog appears, showing you how many locations the action is currently used in. - Click Delete.
Edited Feb 10, 2025 · Erin O'Callaghan
0
Followers
1
Vote
0
Comments